INDEX

Explanations

phrases indicating the uncovering or revelation of information

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate revealing information or providing insight

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

teness

-0.70

icians

-0.67

anwhile

-0.67

heid

-0.65

ãĥĦ

-0.64

zai

-0.64

ierra

-0.63

ONES

-0.62

anamo

-0.61

War

-0.60

POSITIVE LOGITS

brid

0.90

irts

0.83

pload

0.79

 jaws

0.77

iencies

0.77

 hairs

0.76

inished

0.76

ding

0.75

 shed

0.73

 light

0.73

Activations Density 0.019%

phrases indicating the uncovering or revelation of information

phrases that indicate revealing information or providing insight

No Comments

No Known Activations